The historical convergence of population ageing and the emergence of artificial intelligence (AI) technology is driving eldercare service robots from concept to reality. In this process, the scenario becomes the key determinant of whether technology can be translated into service quality. Theoretically, the concept of “scenario” can be abstracted as a form of power relationship, which is constructed by actors and, in turn, constrains their actions. Drawing on Lefebvre’s theory of space, scenarios can be categorized into practical, conceived, and representational ones. These correspond to the service scenarios of industrial robots, traditional AI robots, and embodied AI robots, respectively, reflecting the power relationships within human-machine interactions across different technological characteristics and illustrating the transformation of application scenarios for eldercare service robots. Regarding scenario management, the paradigms adapted to practical and conceived scenarios are “scenario control” and “scenario management”, respectively. Both adhere to a linear management logic. However, when confronted with the complexity of eldercare services, both expose efficacy dilemmas such as operational rigidity, blind spots in meaning interpretation, systemic closure, and the erosion of subjectivity. The technical characteristics of embodied AI robots and their application in the eldercare field make representational scenarios possible, requiring the shift of the management paradigm toward “scenario governance”, that is, transcending linear management logic with a multi-dimensional network governance logic. Its core tenets encompass the construction of open scenarios, facilitation of multi-stakeholder co-creation, promotion of scenario interconnection, optimization of dynamic scenario evaluation and feedback mechanisms, and enhancement of safety supervision, thereby improving the technical efficacy of embodied AI robots in the field of eldercare services.